You know, I don’t know if that’s what “a Western audience” wants. I know that Youtube has made its service steadily worse over time, and also much more resource intensive. Checking Firefox’s Process Manager, it currently lists four Youtube tabs loaded in memory that total up to 1.4GB. And I know a lot of that isn’t the video: when I load a new Youtube tab, it shows the “placeholder” content, gray lines where text will go, for several seconds before anything really displays.

I mean, Bilibili sucks too? I get popups before videos I watch with anime characters, whee. There’s also the whole currently-is-in-Chinese thing, kind of a drawback for an English-speaker, But it does seem to load a bit faster.

Study also suggests that water is wet.

Not true, there is no Springfield, Hawaii, for example.

(gestures at headline)

Too late.

Spotted a typo, it refers to “Apache Foundation’s OnlyOffice,” when that product is OpenOffice (as one can discover by following the link). This is different from the OnlyOffice mentioned earlier in the paragraph.

I have an acquaintance (not sure if qualifies as a friend) who made news for getting fired for, when they were ill, asking an AI to get a quote out of an article for them, which it just out-and-out made up and got them fired. They aren’t stupid, but maybe a bit too trusting. The world lambasted them, which is a huge shame because they do really good work. One strike, you’re out.

So I am not unsympathetic to people who get, in essence, betrayed by AI. (not literally so, because AI is not a person) But yes I think people should be very careful, more careful they they’re being now certainly.

They also should be ready for if it all falls apart, as training costs increase and becomes harder. Already websites like IMDB are putting CAPTCHAs on their site just to browse it. It used to be an accommodating site to scripts that used it to gather information automatically, but those days seem to be over now. Expect to see that more and more, and in the process, the web becomes that much more annoying to use for plain old human readers too.

I could be really cruel right now, I could point out to people how this is chickens coming home to roost, how it’s always been grossly subsidized and underpriced, how the game was always getting users, companies, the country, the whole damn world hooked on these slop machines and then once they’re reliant on them to jack up the price hugely. I could be really mean about this, I could lord it over people, I could point out how they’ve been dupes this whole time and in the process traitors to all of humanity.

I could. i think I will.

It isn’t! But “everything” is a wide net, and destroying it all would be indiscriminate in who it hurts.

To put it into perspective, that’s around 2.3% of the whole US population.

“Shutting everything down” would do untold harm to many low income people.

Thank you!

Don’t I feel like a fool for trying to give you the benefit of the doubt.

“Ever heard of skills,” wow thanks for beginning with a dismissive statement that implies all of this you’re saying is not only true but obvious. This is not the way to respond to people with strong objections.

Your response is really unimpressive. My point is that LLM training, as it now stands, doesn’t seem like it can possibly adapt to an internet that isn’t full of free information ripe for the taking. If people come to rely on LLMs, how will they get the information to keep up with further advancements in, well, anything?

Assuming that’s true, and that’s a BIG assumption… What makes you think that would matter? AI has no interiority; it isn’t a thinking blob, it’s a text generator. Think of it as a fancy Markov chain.

Even if it were true, where in the chain do new principles, new techniques, new concepts enter into it? All these forms of generative AI can do is regurgitate what’s been fed into it. The worst thing you can train an AI from is AI-generated output.

Yes, the age of all programming is over, because no new libraries or languages will ever be invented and LLMs will this always know everything there is to know about coding based on what’s already been written which will never go obsolete.

Honestly, mocking these things is SO EASY.

But, that’s not alternate history, it’s speculative fiction.

If you can’t log into Windows you can’t change its OneDrive settings! What’s more, the user had no idea what was causing the problem, be it OneDrive or something else, until he did that troubleshooting! And, just setting up a new phone shouldn’t make your computer unbootable for any reason! Geez, way to victim-blame there.

I dunno? It sounds very plausible, exactly the kind of thing that Windows would do. I posted about it to Metafilter some time back and no one there seemed to think it couldn’t happen.